Journal of Clinical Immunology

, Volume 31, Issue 4, pp 690–698

Clinical and Immunomodulatory Effects of Celecoxib Plus Interferon-Alpha in Metastatic Renal Cell Carcinoma Patients with COX-2 Tumor Immunostaining


  • Anita Schwandt
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute
  • Jorge A. Garcia
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute
  • Paul Elson
    • Department of Quantitative Health SciencesCleveland Clinic
  • Jeanie Wyckhouse
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute
  • James H. Finke
    • Department of ImmunologyLerner Research Institute, Cleveland Clinic
  • Joanna Ireland
    • Department of ImmunologyLerner Research Institute, Cleveland Clinic
  • Pierre Triozzi
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute
  • Ming Zhou
    • Department of PathologyCleveland Clinic
  • Robert Dreicer
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute
    • Department of Solid Tumor OncologyCleveland Clinic Taussig Cancer Institute

DOI: 10.1007/s10875-011-9530-x

Cite this article as:
Schwandt, A., Garcia, J.A., Elson, P. et al. J Clin Immunol (2011) 31: 690. doi:10.1007/s10875-011-9530-x



Cycloxygenase-2 (COX-2) is an enzyme involved in prostaglandin E2 (PGE2) synthesis associated with higher renal cell carcinoma stage. COX-2 inhibition enhances interferon (IFN-α) anti-tumor immune effects in pre-clinical models. A phase II trial of celecoxib and IFN-α in a targeted population of metastatic renal cell carcinoma patients with maximal COX-2 expression was conducted.


Cytokine-naive metastatic renal cell carcinoma patients with tumors expressing ≥10% maximal COX-2 staining by immunohistochemistry received IFN-α 5 million units daily and celecoxib 400 mg orally twice daily in an open-label, single-arm phase II trial.


There were 3 partial responses among 17 patients (objective response rate 18%; 95% confidence interval, 4–43%). Time to progression was 5.6 months. Increased tumor staining 3+ for COX-2 was associated with increased baseline peripheral blood PGE2 levels, and these patients demonstrated less PGE2 decrease with therapy. Patients with more 3+ COX-2 staining had significantly more CD3+ (p = 0.004) and CD4+ (p = 0.002) IFN-γ T cells at baseline and a significantly greater decrease in these cells with therapy.


Celecoxib plus IFN-α in renal cell carcinoma (RCC) patients with maximally staining COX-2 tumors does not significantly enhance overall response rates over IFN monotherapy.


COX-2-expressing RCC demonstrates inherent immunosuppression. COX-2 inhibition with IFN results in minimal immunomodulation and no augmented clinical activity in RCC.


Interferon-αcelecoxibrenal cell carcinomaimmunotherapyT-regulatory cells


Therapy for metastatic renal cell carcinoma (mRCC) has historically been cytokine-based including interferon-alpha (IFN-α), as immune system manipulation resulted in greater anti-tumor effects when compared with traditional chemotherapeutic agents. IFN-α has produced objective response rates on the order of 10–15% with progression-free survival of approximately 5–6 months. Despite the advent of targeted therapy against vascular endothelial growth factor and mammalian target of rapamycin pathways supplanting IFN-α as first-line therapy, mRCC still remains a largely incurable disease. Exploration of novel mechanisms to enhance host immune cell activation remains a viable investigative endeavor to augment the anti-tumor effect of IFN-α.

There is an expanding list of immune targets and pathways by which tumors evade their host's surveillance systems. To this end, RCC is a prototypical tumor in which a shift from type 1-mediated immune T-cell activity (Th1) to an immunosuppressive type 2 T-cell activity (Th2) predominates [1]. Regulatory T cells (Treg) (CD4+ CD25+ FOXP3+) play a critical inhibitory role in innate anti-tumor responses. An increased number of Tregs in the peripheral blood as well as tumor sites in a variety of tumors supports this paradigm of tumor evasion of host defenses by tilting the balance towards that of a suppressed immune system [24]. Dendritic cells (DCs) detect and are stimulated by foreign antigens, differentiate, and immunologically mature to function as amplifiers to further stimulate effector cell pathways including cytotoxic T cells, natural killer cells, and humoral pathways. Evaluation of mRCC tumors demonstrates a reduction in numbers of infiltrating DCs [5]. Functional evaluation of the decreased number of DCs in mRCC tumors demonstrated an immature and inactivated DC phenotype with impaired antigen-presenting capability [6].

IFN-α is a major cytokine which drives differentiation of the Th1 immune phenotype with stimulation of dendritic cells which promote maturation, differentiation, and activation of cytotoxic T cells [7]. Depletion of CD4+ CD25+ Tregs was shown to enhance the IFN-α-induced anti-tumor immunity in a mouse melanoma model [8]. Thus, increased Tregs as seen in mRCC tumors may alter the anti-tumor activity of IFN-α, and strategies to reduce the number of immunosuppressive Tregs in mRCC may also enhance IFN-α activity. IFN-α may also have anti-angiogenic effects, as it has been associated with decreased tumor growth and vascularization in in vitro studies and murine bladder cancer models [9].

The COX-2 pathway, which results in the conversion of arachadonic acid to prostaglandin E2 (PGE2), may be associated with tumor development by a variety of mechanisms such as neovascularization, inhibition of apoptosis, and stimulation of tumor growth [10, 11]. Furthermore, inhibition of COX-2 has been correlated with a decrease in angiogenesis and tumorigenesis in colon, esophageal, and solid tumors. COX-2 overexpression has been demonstrated in an RCC line (OS-RC-2), and suppression of tumor growth and angiogenesis with COX-2 inhibition was observed in an orthotopic murine model [12]. DCs exposed to PGE2 resulted in increased production of IL-10 and decreased production of IL-12; COX-2 inhibition in these models led to reversal of this phenotype and enhanced DC activity [1315]. In mouse lung cancer models, PGE2 enhanced the inhibitory effects of Tregs via induction of Treg transcription factor FoxP3, and COX-2 inhibition of PGE2 reduced the number and activity of Tregs and decreased tumor burden [16].

Celecoxib is a nonsteroidal anti-inflammatory agent which is a selective COX-2 inhibitor. A previously conducted prospective study of COX-2 non-selected mRCC patients evaluating the activity of celecoxib 400 mg BID administered in conjunction with IFN-α 3 MU daily demonstrated 20–30% of mRCC tumors with maximal (3+) staining for COX-2 [17]. There was a significant association between both objective response rate (ORR) and progression-free survival (PFS) with higher intensity and percentage of COX-2 staining. Similarly, all objective responses were confined to the subgroup of patients with at least 10% maximal COX-2 tumor immunostaining. Intriguingly, increased COX-2 expression was associated with adverse prognostic risk factors and higher tumor grade. However, despite these adverse prognostic features, PFS was longer in mRCC with maximal COX-2 expression treated on this clinical trial. Based on the above considerations, a phase II trial evaluating the combination therapy of celecoxib and IFN-α in a targeted population of mRCC patients prescreened for maximal COX-2 expression was conducted to evaluate the clinical and immunomodulatory effects.

Materials and Methods

Patient Selection

Between May 2006 and July 2009, patients with metastatic RCC were enrolled in this prospective, open-label, single-arm trial with IFN-α and celecoxib. The treatment regimen, monitoring, and laboratory correlate studies were approved by the institutional review board of the participating institution and all patients signed a written, informed consent.

Eligible patients had histologically confirmed metastatic clear cell RCC with maximal COX-2 overexpression in ≥10% of the tumor (see methods below). Patients had measurable disease by Response Evaluation Criteria in Solid Tumors (RECIST) criteria [18]. Eastern Cooperative Oncology Group (ECOG) performance status of 0–2 and normal organ and marrow function defined by the following: leukocytes ≥3,000/μL, absolute neutrophil count ≥1,500/μL, platelets ≥75,000 ≥ μL, total bilirubin ≤1.5× institutional upper limit (IUL) of normal, alanine aminotransferase and aspartate aminotransferase ≤2.5× IUL of normal, and creatinine clearance ≤2.0 IUL of normal.

Patients must not have received prior cytokine therapy for RCC. The number of prior non-cytokine therapies was not limited. Exclusion criteria included significant cardiovascular disease including congestive heart failure (New York Heart Association class III or IV), active angina pectoris requiring nitrate therapy, uncontrolled dysrhythmias, or recent cardiovascular event (defined as any of the following within the previous 6 months: TIA/CVA, MI, vascular surgery). Patients with treated, clinically stable central nervous system metastases were permitted; however, since corticosteroid treatment can decrease COX-2 expression and reduce the immune response to IFN-α, concurrent systemic steroid therapy was prohibited. Patients with a history of a severe allergic reaction to sulfonamide or derivatives were excluded. Women were excluded if pregnant or breastfeeding and were required to have a negative pregnancy test with the use of adequate contraception while on study. Patients’ baseline risk assessment was performed by Memorial Sloan Kettering Clinical Criteria (MSKCC) risk criteria [19].

Treatment Plan

Patients with mRCC meeting eligibility criteria received IFN-α 5 MU s.c. 5×/week in combination with celecoxib 400 mg p.o. twice daily continuously in 28-day cycles. Patients underwent treatment assessment with CT scans and bone scan at baseline, at week 4 of cycle 2, and every other cycle thereafter. Patients were clinically evaluated every 4 weeks with a physical exam and toxicity assessment. Patients remained on study until progressive disease as defined by RECIST criteria, comorbid illness precluding further treatment, unacceptable adverse event, or patient withdrawal of consent.

Toxicities were graded according to the CTEP Common Toxicity Criteria version 3.0. Any grade 3 or 4 toxicity attributable to therapy required IFN-α and celecoxib to be held. If toxicity resolved to ≤ grade 1 within 4 weeks, IFN-α was reinstituted at 1 MU less than the prior dose and celecoxib was reinstituted at 400 mg BID. If the toxicity resolved to ≤ grade 1 within 2 weeks, the IFN-α dose could be left at the prior dose according to the discretion of the treating physician.

Tumor Sample Evaluation

Paraffin-embedded tumor samples from prior nephrectomy or tumor biopsy were retrieved. In addition to pathological verification of clear cell histology, tumor tissue was stained for COX-2 to determine eligibility for the trial. The percentage of tumor cells with none, 1+, 2+, and 3+ COX-2 staining was recorded. Tumor cells were also co-stained for T cells (CD4, CD8, CD25) and for FOXP3.

Immunohistochemical stains were performed using the standard avidin–biotin complex system on a Vantana XT workstation. Briefly, 5-μm tissue sections were prepared from representative blocks of each case and deparaffinized. Antigen retrieval was performed in Tris/borate/EDTA buffer, pH 8. The sections were incubated sequentially with primary antibodies, biotinylated secondary antibodies, streptavidin conjugated with horse radish peroxidase, and chromogenic substrate DAB. The source and dilution of primary antibodies were as follows: COX-2: Biosciences Pharmingen, 1:20; CD4: Novocastra, 1:10; CD8: Dako, 1:50; and CD25: Novocastra, 1:10. Anti-COX-2-stained slides were reviewed and scored for staining intensity (on a scale of 0 for no stain to 3+ for intense complete membrane and cytoplasmic staining). Percent of tumor cells staining at any degree of intensity was recorded along with stromal and normal kidney tissue staining. For each of the markers of immunoregulatory cells (CD4, CD8, FOXP3), 20 high power (×400) fields were examined for each tumor and the average positive cells per high power field were calculated. For COX-2, the positive control tissue used for immunostaining was a breast tumor with known overexpression of COX-2 enzyme by immunohistochemistry, western blot, and RT-PCR. Internal positive control was adjacent normal kidney, and internal negative control was tumor stroma and stromal blood vessels. For CD4 and CD8, tonsil was used as positive control tissue.

Peripheral Blood Assays

Correlative study blood collection was obtained at baseline for all screened patients and at the following time points for all patients who were enrolled in the study: day 1 of cycle 2 and day 1 of cycle 3 (and at off-study time point). Seven × 10 cc EDTA tubes of peripheral blood were collected, subjected to Ficoll-Hypaque centrifugation to isolate peripheral blood mononuclear cells (PBMCs) which were then stored in liquid nitrogen.

PGE2 Levels

PGE2 and metabolites were assayed for measurement of adequacy of COX-2 inhibition. Because in vivo PGE2 is rapidly converted to an inactive metabolite (13,14-dihydro-15-keto PGE2) by the prostaglandin 15-dehydrogenase pathway, the half-life of PGE2 in the circulatory system is approximately 30 s and normal plasma levels are 3–12 pg/mL. Therefore, the determination of in vivo PGE2 biosynthesis is often best accomplished by the measurement of PGE2 metabolites. A commercially available Prostaglandin E Metabolite assay (Cayman Chemical, Ann Arbor, MI, USA) was employed that converts all major PGE2 metabolites into a single stable derivative which is easily measurable by enzyme immunoassay.

Th1/Th2 Phenotyping

Viable PBMCs (1 × 106) were stimulated for 72 h with anti-CD3 plus anti-CD28-coated Dynabeads in RMPI-1640 containing 10% FCS and then GolgiPlug (BD Biosciences) was added for the last 6 h to prevent cytokine secretion. Cells were labeled with APC-anti-CD3 Ab (BD) to define the T-cell population or with PerCP-anti-CD4 (BD) to detect the major T-helper subset. Cells were fixed and permeabilized and subsequently stained with antibodies to FITC-IFN-gamma (BD). Cells were acquired on BD FACSCalibur and analysis was performed using FlowJo software.

T-regulatory Cells

Treg cells were detected by surface staining PBMC with anti-CD3 (APC), anti-CD4 (PerCP), and anti-CD25 (PE) (Stem Cell Technologies) followed by cell permeabilization and then staining with anti-Foxp3 antibody (AF-488, eBioscience). Non-stained cells served as a negative control. Data were acquired using Cell Quest on a BD FACSCalibur and analyzed using FlowJo software. At least 300,000 live cell events were collected for each tube used in the analysis.

DC Isolation

DCs were isolated from peripheral blood using the Blood Dendritic Cell Isolation Kit II, a magnetic labeling system for concurrent isolation of plasmacytoid and myeloid dendritic cells from PBMCs. Isolation was performed in a two-step MACS® separation: First, B cells and monocytes were magnetically labeled and depleted using a cocktail of CD19 and CD14 MicroBeads. Subsequently, the pre-enriched dendritic cells in the non-magnetic flow through fraction were magnetically labeled and enriched using a cocktail of antibodies against dendritic cells markers BDCA-4 (Neuropilin-1), BDCA-3, and CD1c (BDCA-1). The highly pure enriched cell fraction comprised BDCA-2+ plasmacytoid dendritic cells and two subsets of myeloid dendritic cells, CD1c (BDCA-1)+ MDC1 and CD1c (BDCA-1) BDCA-3bright MDC2. B cells and monocytes were depleted in advance because a subpopulation of B cells expresses CD1c (BDCA-1), and monocytes express BDCA-3 at low levels. The kit included FcR Blocking Reagent, the Non-Dendritic Cell Depletion Cocktail, and the Dendritic Cell Enrichment Cocktail.

DC Activation Status

A quantitative real-time polymerase chain reaction (QRT-PCR) method to assess the activation status of DC was used which compared the relative production of IL-12 and IL-10 messenger RNA (mRNA). IL-12 and IL-10 protein production, as determined by ELISA, parallels IL-12p40 and IL-10 mRNA production, as determined by QRT-PCR. Furthermore, the IL-12/IL-10 mRNA ratio, as determined by QRT-PCR, correlates with the activation status, as determined by MLR. A DC activation index was calculated for each sample by multiplying the number of DC per milliliter of blood by the ratio of IL-12/IL-10 mRNA expressed.

Statistical Analysis

The sample size for this trial was based on the primary endpoint, ORR. Objective response was assessed by the RECIST criteria. The null hypothesis was that the objective response rate is <20%. An increase in the underlying response rate to 40% or greater would be considered significant and worthy of further study. The study employed a two-stage accrual design with interim analysis after the initial treatment of 17 patients. If 4 or more patients showed an objective response, accrual would be expanded to a total of 34 patients. If 10 or more of the 34 patients responded (observed response rate >29%), the combination would be accepted for further testing. With this design, the overall type I and II errors were 0.11 and 0.10, respectively, and the likelihoods of stopping early under the null and alternative hypotheses were 0.55 and 0.05, respectively. Time to progression (TTP) was measured from the start of treatment to the date of documented progression or death, whichever came first, and was summarized using the method of Kaplan and Meier.

Immunologic and outcome data were analyzed primarily using semi- and non-parametric methods. Changes from baseline were analyzed using the Wilcoxon signed rank test, and associations with COX-2 expression were analyzed using the Jonckheere–Terpstra test, the Wilcoxon rank sum test, Fisher’s exact test, and Spearman rank correlations. The Cox proportional hazards model was used to assess TTP.

All tests of statistical significance were two-sided and p values ≤ 0.05 were considered statistically significant. All analyses were performed using SAS version 8 (SAS Inc., Cary, NC, USA) and StatXAct 7 (Cytel Inc., Cambridge, MA, USA).


Patient Characteristics

Between May 2006 and July 2009, a total of 66 patients who gave consent had screening analysis of the amount and intensity of COX-2 expression of their RCC tumor specimens determined. The study was discontinued after a planned interim analysis at 17 patients did not achieve the predefined minimum ORR. Of the 66 patients whose tumors were screened for COX-2 expression, 37 patients (56%) exhibited ≥10% maximal COX-2 expression (3+). Table I shows the baseline characteristics of the 17 patients with maximal COX-2 expression who ultimately enrolled on trial. Other patients with maximal COX-2 expression were not treated on trial due to ineligibility or patient/physician preference for alternative treatment.
Table I

Patient characteristics (n = 17)


Patients (%)

Median age

64 years (range 43–77)




13 (76%)


4 (24%)

ECOG performance status



11 (65%)


6 (35%)

Prior nephrectomy

17 (100%)

Prior radiotherapy

3 (18%)

Prior systemic therapya

6 (35%)


3 (18%)


4 (24%)


2 (12%)


1 (6%)

Number of prior systemic regimens



11 (65%)


2 (12%)


3 (18%)


1 (6%)

Metastatic disease



12 (71%)

 Lymph nodes

7 (41%)


2 (12%)


3 (18%)


14 (82%)

aThe total number of patients treated with each agent totals >6 because patients received multiple therapies

The patient characteristics of this study cohort were typical of a phase II metastatic RCC population. The median age was 63 years and patients were predominantly male. Thirty-five percent of patients had prior non-cytokine therapy, consisting of sunitinib, sorafenib, axitinib, or bevacizumab. The majority of patients had not received prior systemic therapy. The majority of patients had metastatic disease to the lungs (71%) and lymph nodes (41%). Patient risk stratification category by MSKCC criteria was 18% favorable, 82% intermediate, and 0% high risk.

Treatment Administration

Patients received a median of 4 cycles of therapy (range 1–20 cycles), with 11 patients (65%) receiving at least 3 treatment cycles. Six patients (35%) required dose reductions of IFN-α. At the time of this analysis, all patients had discontinued therapy. Most patients discontinued therapy secondary to PD (65%), three patients (18%) discontinued for adverse events, one patient (6%) died while on trial (related to disease progression), one patient discontinued for unspecified reasons, and one patient withdrew consent shortly after initiation of therapy.

Clinical Outcome

Three patients achieved a RECIST defined objective partial response for an ORR of 18%. Five patients (29%) had the best response of SD which was maintained over a range of 2.6–12.9 months, while 41% of patients had a best response of progressive disease. Two patients (12%) discontinued the trial prior to first imaging evaluation. One patient withdrew consent and one patient discontinued for adverse events. Of the 15 patients who were evaluable after cycle 2 with imaging, a median 1.6% decrease in overall tumor burden (range 75% decrease to 33% increase) was demonstrated. Nine of 15 patients (60%) experienced some degree of reduction of tumor burden (median 11.6%, range 0.6–75%). The median TTP was 5.6 months. Median overall survival was 14.4 months.


Toxicity was consistent with prior clinical experience with single agent IFN-α. Fatigue (82%), nausea and vomiting (53%), anorexia (53%), diarrhea (41%), skin changes (41%), and depression (29%) were the most common reported side effects. Laboratory toxicities (all grades) included thrombocytopenia (59%), lymphopenia (53%), leukopenia (47%), anemia (41%), and neutropenia (35%). IFN-α dose reductions were required in 35% (6/17) of patients overall and 55% of the 11 patients on study for 3 or more cycles.

Baseline Immune Parameters and Treatment Modulation

Consistent with prior studies, mRCC patients as compared with normal patients had decreased T-cell IFN-γ production and increased numbers of Tregs (data not shown). Data regarding tumor COX-2 expression and immune parameters were analyzed using several arbitrary codings of COX-2 staining, including (a) the proportion of cells showing 1+ or greater COX-2 tumor staining; (b) the proportion of cells showing 3+ staining; (c) no 3+ COX-2 staining, vs. 3+ staining in <10% of cells vs. 3+ staining in ≥10% of cells; and (d) because maximal COX-2 overexpression in ≥10% of the tumor was an eligibility criterion, 3+ staining in <20% vs. ≥20% of cells. Only two patients had no COX-2 tumor expression of any degree, and therefore, comparisons of this group to the COX-2-expressing group were not considered.

Tumor Immunostaining

Analyzing all screened patients (n = 66), there were no correlations between intra- or peri-tumoral CD4, CD8, or FOXP3 staining and the percent of cells staining 3+ for COX-2 regardless of how COX-2 staining was coded (all p ≥ 0.10). There was, however, an association between the total proportion of cells showing at least 1+ COX-2 staining and both intra- and peri-tumoral CD4+ T cells (r = 0.39, p = 0.02 and r = 0.36, p = 0.02; Fig. 1a, b) and intra-tumoral CD8+ T cells (r = 0.33, p = 0.04; data not shown). As with 3+ COX-2 staining, there was no indication of an association with peri-tumoral CD8+ T cells or FOXP3+ Tregs (all p ≥ 0.12).
Fig. 1

The percentage of tumor cells staining positive for COX-2 were plotted against the average number of cells that stained positive for anti-CD4 antibody by immunohistochemistry. An association between the total proportion of cells showing at least 1+ COX-2 staining and the number of both intra- and peri-tumoral CD4+ T cells was observed (r = 0.39, p = 0.02 and r = 0.36, p = 0.02; a, b). Similar findings were observed between COX-2 staining and intra-tumoral CD8+ T-cell staining (r = 0.33, p = 0.04)

Peripheral Blood Parameters

Prostaglandin E2

Considering all screened patients, an increased percentage of tumor staining 3+ for COX-2 expression was associated with increased baseline peripheral blood PGE2 levels (none vs. >0–10% vs. ≥10% 3+ COX-2 expression (p = 0.02)). However, there was no association between PGE2 levels and the overall proportion of cells with any COX-2 staining (p = 0.40). Considering only the patients who entered the clinical trial, patients with <20% vs. ≥20% 3+ COX-2 tumor immunostaining exhibited similar changes in PGE2 levels following 1 cycle of therapy (median 0.23 and 0.25 ng/mL increases, respectively, p = 0.70). However, after 2 cycles of treatment, patients with <20% 3+ COX-2 staining had a significantly greater decrease in PGE2 levels compared to patients with ≥20% 3+ staining (median 0.81 ng/mL decrease vs. 0.25 ng/mL increase; p = 0.03; Fig. 2).
Fig. 2

The changes in serum PGE2 metabolite levels were defined after 2 cycles of treatment (IFN-α/celecoxib) and plotted against the number of tumors staining 3+ (>20% of cells) or not for COX-2. As noted, patients with <20% 3+ COX-2 staining had a significantly greater decrease in PGE2 levels compared to patients with ≥20% 3+ staining (median 0.81 ng/mL decrease vs. 0.25 ng/mL increase; p = 0.03)

T-Cell Parameters

There was no significant association between CD3+ or CD4+ IFN-γ-producing T cells at baseline and the overall proportion of cells staining positive for COX-2 among the patients screened for the trial (p ≥ 0.12 in both cases). There was, however, some suggestion of a positive correlation of these cells with the proportion of cells with maximal COX-2 staining (r = 0.27, p = 0.08 and r = 0.31, p = 0.04, respectively), with patients in whom ≥20% of cells were 3+ having significantly higher numbers of CD3+ and CD4+ IFN-γ-producing T cells (median 13.5 and 12.5, respectively) than patients with <20% of cells staining 3+ (median 5.5 and 4.0, respectively), p = 0.004 and 0.002, respectively (Fig. 3). A similar association was observed restricting the analysis to patients who entered the clinical trial (data not shown).
Fig. 3

The proportion of IFN-γ+ CD3+ (a) and CD4+ (b) T cells at baseline was plotted for patients demonstrating maximal (3+) COX-2 staining in <20% vs. ≥20% of tumor cells. Patients in whom ≥20% of cells were 3+ COX-2 had significantly higher numbers of CD3+ and CD4+ IFN-γ-producing T cells (median 13.5 and 12.5, respectively) than patients with <20% of cells staining 3+ (median 5.5 and 4.0, respectively), p = 0.004 and 0.002, respectively. For convenience, a horizontal line is drawn at 10% to illustrate the difference between the two groups

Overall, patients on the clinical trial did not demonstrate statistically significant changes in these T-cell populations (IFN-γ+) following 1 (p = 0.67 and 0.63, respectively) and 2 (p = 0.23 and 0.34, respectively) cycles of therapy, although the counts tended to decrease (median decreases of 14% and 13% after 1 cycle of therapy and 23% and 26% after 2 cycles, respectively). The changes, however, were negatively correlated with the proportion of cells with 3+ COX-2 immunostaining. That is, patients with more extensive 3+ staining tended to exhibit greater relative decreases in both CD3+ or CD4+ IFN-γ-producing T cells than patients with less extensive 3+ staining (all r < −0.60 and all p < 0.02). An example of the correlations is given in Fig. 4, which plots changes in CD3+ IFN-γ-producing T cells after 2 cycles of therapy as a function of the proportion of tumor cells with maximal COX-2 staining (r = 0.92; p < 0.001). This result, in conjunction with the observation that patients with limited 3+ COX-2 staining (<20% of cells) tended to have a greater decrease of PGE2 levels following 2 cycles of therapy (Fig. 2), suggests that although PGE2 depletion with celecoxib was unable to significantly augment the IFN-γ T-cell population in patients with COX-2-expressing RCC tumors, the lack of effect was greater in patients with more extensive 3+ staining. Three of 4 patients with <20% of cells staining 3+ had a decline in PGE2, compared to pretreatment levels, which was accompanied by a modest or no decrease in CD3+ IFN-γ-producing T cells. In contrast, PGE2 levels increased in all 4 patients in whom ≥20% of cells were 3+ COX-2, and this was associated with 53–95% decreases in CD3+ IFN-γ-producing T cells. There were no significant associations between baseline nor change in peripheral blood Tregs and COX-2 staining (data not shown).
Fig. 4

The percentage of pretreatment tissue that stained 3+ for COX-2 was plotted against the percentage change in T cells (CD3+) producing IFN-γ after 2 cycles of treatment with interferon and celecoxib. Intracellular levels of IFN-γ were measured in T cells after in vitro stimulation with anti-CD3/anti-CD28. There was a negative correlation between the degree of COX-2 staining of tumors and the percentage of peripheral blood T cells producing IFN-γ after 2 cycles (r = 0.92, p < 0.001)

Dendritic Cell Parameters

Although there was no association between peripheral blood DC numbers and the overall proportion of tumor cells staining positive for COX-2 in the cohort of screened patients (p = 0.39) or the proportion of cells staining 3+ for COX-2 (p = 0.56), there was some indication of a negative correlation between the proportion of 3+ COX-2 tumor cells and baseline DC function as measured by DC IL-12 production (r = −0.35, p = 0.05), but not IL-10 production (p = 0.98). Consequently, more 3+ COX-2 expression negatively correlated with the IL-12/IL-10 ratio (r = −0.34, p = 0.06). There was no association between COX-2 staining and changes in DC IL-12 or DC IL-10 with treatment (data not shown).

Objective Response, COX-2 Staining, and Immune Parameters

Neither objective response nor TTP was associated with any of the immune parameters examined. Patients whose best objective response was PR or stable disease, however, tended to have more extensive 3+ COX-2 immunostaining (median 22%) than patients who progressed (median 10%; p = 0.07), with 5/6 (83%) patients having ≥20% of tumor cells staining 3+ responding or having stable disease vs. 3/9 patients (33%) with <20% 3+ staining (p = 0.12).


Historically, cytokine therapy has been the standard of care in the treatment of mRCC. While overall response rates with cytokine therapy are modest, in select patients, cytokines act in concert with the host immune system to induce durable complete responses. Consequently, immunotherapy remains a viable option for further investigation. Efforts to clarify the heterogeneity of response of mRCC to immune therapy and to identify agents which act synergistically with interferon to augment the immune and clinical response are rational investigative endeavors.

The relationship between RCC tumor COX-2 expression and outcome has been reported, although the precise nature of the relationship is unclear. One prior report noted an enhanced 5-year survival after nephrectomy in COX-2 negative patients, although the vast majority of patients had non-metastatic disease [20]. A more recent study in metastatic RCC patients predicted a longer median overall survival in patients with higher levels of COX-2 immunostaining [21]. These discordant results highlight the inherent variability and subjectivity of immunostaining as well as issues of tissue sampling, type of tissue analyzed (nephrectomy sample, metastatic tissue), timing of tissue acquisition relative to metastases, etc. A prior study suggested activity of the combination of IFN-α and celecoxib COX-2 inhibition in metastatic RCC patients with maximally expressing COX-2 tumors [17]. Despite preselection for maximal COX-2 expression, this prospective study of a combination of two agents did not show significantly enhanced clinical outcome compared to historical response rates of interferon monotherapy [19]. It is noteworthy that inclusion criteria allowed for multiple lines of prior non-cytokine therapy. Indeed, 35% of patients had prior treatment with a targeted agent such as sunitinib, sorafenib, axitinib, or bevacizumab. Thus, some of the enrolled patients were beyond first-line therapy and decreased clinical effect may be expected when compared to first-line IFN-α. Further, the vast majority of patients were MSKCC intermediate risk category, with data from a separate trial suggesting minimal effect of cytokines in this risk group [22]. It is likely that the anti-tumor effect of interferon, either alone or with additional immune system manipulation, requires restricted application to exclusively good risk patients.

Nonetheless, significant and favorable immune modulation was not consistently observed in this study. Notably, peripheral blood PGE2 levels actually increased in patients with greater COX-2 tumor expression. This would be the opposite of the desired effect to lower PGE2 levels and favorably alter the inherent immunosuppressive phenotype of these tumors. These data support that PGE2 regulation in RCC patients is complex and requires either more COX-2 inhibition than can be achieved with the celecoxib dose and schedule used in this trial, or perhaps alternative mechanism(s) of inhibition. Paradoxically, patients with greater tumor COX-2 staining had more intra- and peri-tumoral T cells and more IFN-γ-producing T cells at baseline. These quantitative advantages clearly did not result in heightened clinical effect of IFN-α immunotherapy. It can be hypothesized that the IFN-γ T-cell response was qualitatively insufficient to stimulate an anti-tumor T-cell response. In addition, such patients had greater declines in these cell populations with therapy. Thus, the COX-2-expressing tumor environment has a quantitatively and likely qualitative detrimental effect on IFN-γ-producing T cells that is not overcome with PGE2 depletion by celecoxib. In addition, we were unable to demonstrate an effect of therapy on the Treg subpopulation. While the number of CD4+ CD25+ Treg was not directly evaluated in the mRCC tumor microenvironment during therapy, peripheral blood levels of Tregs may indirectly reflect tumor levels and were not favorably altered. Thus, the lack of effect on Tregs may have also contributed to the lack of more favorable immune modulation and clinical effect observed in this study. Further, although DC numbers were maintained in high COX-2 expressors, a favorable altering of the IL-12/IL-10 ratio with treatment could not be demonstrated, again underscoring the limited immunomodulation observed with therapy that likely impacted the clinical effects seen.

This study has several limitations. Measurements of immune parameters occurred at prespecified intervals including screening, baseline, and after each cycle. These snapshots in treatment period may overlook potential initial changes which may have occurred within the first few weeks of treatment. Our first analysis of peripheral blood after initiation of IFN-α and celecoxib at 4 weeks showed no significant change in Tregs. However, a study of IFN-α in mRCC analyzed Treg cells at an earlier time point and measured significantly decreased numbers of CD4(+) and FoxP3(+) Treg cells 2 weeks after initiation of IFN-α, which subsequently increased again as treatment continued [23]. Nonetheless, non-sustained decreases in Treg are unlikely to significantly impact the clinical effect of immunotherapy. Additional efforts at more significant and sustained effects will be needed to augment the clinical benefits of immunotherapy.


This study was supported by NIH funding (R21CA123854 and R21CA123954). Funding for this study was provided by Merck & Co. Inc, Whitehouse Station, NJ (formerly Schering Plough Corp) and Pfizer Oncology (New York, NY).

Copyright information

© Springer Science+Business Media, LLC 2011